determines weights according to the preference of decision makers or experts, which is
volatile due to the unreliability judgment of decision makers. The representative methods
of subjective approaches are analytic hierarchy process (AHP) [14] and fuzzy MADM [15].
However, the fuzzy MADM methods require cumbersome computations, and they are not
suitable for solving problems with more than ten alternatives [16]. The objective ap-
proaches rely on the performance of the evaluation objects to determine the weight,
whereas, it does not consider the subjective idea of decision-makers. The representative
methods of objective approach are entropy [16] and principal component analysis (PCA).
While, PCA requires the meaning of the weights should be exclusive statistical and it also
may lead to information lost.
Consideration of the user preference, and restraining the subjectivity and arbitrariness,
the subjective and objective approaches should be combined. There are numbers of pub-
lications that combined two approaches together to determine the weights. Reference [17]
combined AHP with entropy to evaluate the safety of smart grid. In [18], the authors
proposed an improved fuzzy comprehensive evaluation method for assessing the perfor-
mance of leak detection systems, while the weights are calculated by combing entropy with
AHP. Reference [19] used the combined weight, i.e. AHP and entropy methods, to de-
termine the importance of the evaluation criteria. Reference [20] used AHP and entropy to
acquire attribute weights’ vectors, and then synthesized the weights’ vectors by using
group decision making.
After the weights have been determined, the alternatives (i.e. number of vehicles)
should be ranked before obtaining the optimal network performance of VANETs.
Several evaluation methods have been used for network selection to rank the alterna-
tives in heterogeneous wireless networks, such as technique for order preference by
similarity to ideal solution (TOPSIS) [21]. The authors developed weighted rating of
multiple attributes (WRMA) method to rate attribute, and TOPSIS was employed to rank
networks in Ref. [22]. However, Ref. [23] pointed out that grey relational analysis (GRA)
achieved better performance than TOPSIS for specific service type, i.e. interactive and
background. The author in [24] applied entropy and TOPSIS for network selection.
Due to the fact that in VANETs, the safety-related application that is selected by the
user is subjective and the network performance metrics are objective, we use a hybrid
approach, which combines AHP with entropy approaches, to evaluate the performance
comprehensively. Then, GRA [25] is used for ranking. And it has been used to provide a
simple and transparent evaluation procedure for a clear cut ranking order of strategies
derives in [26].
In this paper, a comprehensive evaluation scheme has been proposed to obtain the
optimal performance of VANET. The AHP and entropy approaches have been combined to
determine weights. The former determines the relative weights among different ACs ac-
cording to the user preference, i.e. safety-related applications. And the latter is used to
determine the weights of performance metrics, i.e. time delay, throughput and PDR. Then
the GRA has been adopted to rank the alternatives to achieve the highest resource uti-
lization of VANETs. To the authors’ knowledge, this is the first time to determine the
weights among different ACs in VANETs, which takes the user preference into account.
The main contributions of this paper are summarized as follows: (1) a comprehensive
scheme is proposed for the performance evaluation of VANETs. (2) The proposed scheme
takes the user preference and the performance metrics into account. And it uses AHP and
entropy to calculate the weights. (3) The comprehensive evaluation results are ranked using
GRA to obtain the optimum network performance.
Comprehensive Performance Evaluation of VANETs for… 3019
123